Estimating willingness to pay for public health insurance while accounting for protest responses: A further step towards universal health coverage in Tunisia?

Introduction Developing countries face major challenges in implementing universal health coverage (UHC): a widespread informal sector, general discontent with rising economic insecurity and inequality and the rollback of state and public welfare. Under such conditions, estimating the demand for a health insurance scheme (HIS) on voluntary basis can be of interest to accelerate the progress of UHC-oriented reforms. However, a major challenge that needs to be addressed in such context is related to protest attitudes that may reflect, inter alia, a null valuation of the expected utility or unexpressed demand. Methods We propose to tackle this by applying a contingent valuation survey to a non-healthcare-covered Tunisian sample vis-a-vis joining and paying for a formal HIS. Our design pays particular attention to identifying the nature of the willingness-to-pay (WTP) values obtained, distinguishing genuine null values from protest values. To correct for potential selection issues arising from protest answers, we estimate an ordered-Probit-selection model and compare it with the standard Tobit and Heckman sample selection models. Results Our results support the presence of self-selection and, by predicting protesters' WTP, allow the true sample mean WTP to be computed. This appears to be about 14% higher than the elicited mean WTP. Conclusion The WTP of the poorest non-covered respondents represents about one and a half times the current contributions of the poorest formal sector enrolees, suggesting that voluntary participation in the formal HIS is feasible.

Automated summary

This study proposes to address the issue of protest attitudes in estimating the demand for a health insurance scheme in Tunisia by conducting a contingent valuation survey, distinguishing genuine null values from protest values and correcting potential selection issues. The results support the presence of self-selection and enable the computation of the true sample mean willingness-to-pay.